Preservation of global order equivalence

Psychological data frequently can only be gives rank-order significance. This paper investigates and characterizes those cluster methods suitable for use in such a situation. The work is carried out in an abstract order theoretic setting but a relation is shown to earlier models due to Jardine and Sibson as well as Matula. It deals with that portion of the clustering process that starts with a dissimilarity measure defined on a set P and produces a nested sequence of reflexive symmetric relations on P.